[CMake] Status/use of CMake on Cygwin

marco atzeri marco.atzeri at gmail.com
Mon May 20 01:51:42 EDT 2013


Il 5/20/2013 3:28 AM, Alan W. Irwin ha scritto:
> On 2013-05-19 21:06-0000 David Cole wrote:
>

> With regard to your frustration with Cygwin speed, I must say I have
> similar frustrations with how fast commands execute under bash.exe.
> That issue appears to be
> entirely due to startup latency.  For example, here is a comparison of
> "cmake --version" results on Linux versus running that same command
> under bash.exe on Wine.
>
> software at raven> time cmake --version
> cmake version 2.8.10.2
>
> real    0m0.008s
> user    0m0.000s
> sys     0m0.004s
>
> bash.exe-3.1$ time cmake --version
> cmake version 2.8.10.2
>
> real    0m0.197s
> user    0m0.000s
> sys     0m0.040s
>
> These results were determined with a CMake version I built myself on
> Linux and a downloaded Windows binary (built by Kitware with Microsoft
> proprietary compilers) on MinGW/MSYS/Wine, but in the past I have also
> gotten similar bad results for a CMake version that I built myself on
> MinGW/MSYS/Wine, and there are similar bad startup latency results
> (0.2 seconds just to execute the --version option for the command
> which is a factor of ~25 slower than the corresponding Linux result)
> for all other heavily used build commands such as gcc and make.  So
> this is a general bash.exe issue and not a cmake issue, and it a real
> killer of build speeds since cmake and gcc are often invoked during
> the build process (e.g., to provide progress reports, to make a quick
> build test, or to actually make a build containing many separate small
> source files ) in a way that is completely dominated by the bad
> startup latency numbers above.
>
> At this point I don't know whether this startup latency issue is due
> to an issue with bash.exe on Wine or a more general bash.exe issue on
> Windows, and if anybody hear could do such a simple "time cmake
> --version" test for a MSYS/bash.exe environment on Microsoft Windows
> to see whether the result is nearer 0.200 seconds or 25 times faster
> that might confirm or eliminate Wine as the culprit.

the slowness at startup on cywin is mainly due the fork
implementation
http://cygwin.com/faq/faq-nochunks.html#faq.api.fork

unfortunately windows makes a real fork implementation very inefficient.


>
> Alan
> __________________________
> Alan W. Irwin

Marco



More information about the CMake mailing list